Skip to main content

Table 2 Classification accuracies for unlabeled graph datasets

From: Sequence-to-sequence modeling for graph representation learning

Datasets COLLAB IMDB- BINARY IMDB- MULTI REDDIT- BINARY REDDIT- MULTI-5k REDDIT- MULTI-12k  
# Graphs 5000 1000 1500 2000 5000 11929  
Avg. Nodes 74.49 19.73 13.00 429.61 508.50 391.40  
# Classes 3 2 3 2 5 11 AVG Rank
node2vec (G+16) 56.06 ±0.2 50.17 ±0.9 36.02 ±0.7 71.31 ±2.2 33.11 ±1.7 23.62 ±0.3 6.0
DGK (YW15) 73.0 ±0.2 66.9 ±0.5 44.5 ±0.5 78.0 ±0.3 41.2 ±0.1 32.2 ±0.1 4.5
PSCN (N+16) 72.6 ±2.1 71.0 ±2.2 45.2 ±2.8 86.3 ±1.5 49.1 ±0.7 41.3 ±0.4 3.6
WL-OA (K+16) 80.7 ±0.1 89.3 ±0.3
GCN (KI+17) 80.17 ±1.6 73.6 ±3.4 51.3 ±3.8 72.8 ±2.5 40.6 ±1.8 32.4 ±1.9 3.6
LWL (M+17) 73.5 ±0.5 81.3 ±0.5
DGCNN (Z+18) 73.76 ±0.49 70.03 ±0.86 47.83 ±0.85
SGR (T+18) 71.98 70.38 47.97 87.45 53.22
S2S-N2N-PP 81.75 ±0.8 73.8 ±0.7 51.19 ±0.5 86.50 ±0.8 52.28 ±0.5 42.47 ±0.1 2.1
supervised 8 2 . 3 2 ± 0 . 8 7 4 . 9 ± 0 . 8 5 3 . 2 6 ± 0 . 7 89.10 ±0.8 5 3 . 3 8 ± 0 . 6 43.57 ±0.6 1.0